Child Opportunity Index Release Notes

The first release of the Child Opportunity Index (COI) 3.0 (published in 2024) included data covering the years from 2012 through 2021. The current release, COI 3.0-2023 (published in 2025), extends data coverage through 2023. To distinguish between different releases, each COI 3.0 release is labeled with a suffix indicating the most recent year of data coverage. The first release is now retroactively relabeled as COI 3.0-2021. The next planned release, COI 3.0-2024, is anticipated for 2026 and will add data for 2024.

It is important to note that each release introduces small changes that alter previously published COI 3.0 data. For this reason, COI 3.0 data is not strictly comparable across releases. Each new release includes data from all previously published years and adds more recent data, all of which is processed using the same release-specific protocol. While COI data is not strictly comparable across releases, the correlation between them is very high. For example, the bivariate correlation (Pearson’s rho) between the 2021 overall index composite z-scores (nationally normed) for COI 3.0-2021 and COI 3.0-2023 is 0.99. 

Why different COI 3.0 releases?

First, new releases are necessary to keep the COI current and helpful for many applications that require the most recent data. The new releases of the COI reflect more current data. They also respond to ongoing changes in the underlying data sources, which affect current and historical data. The COI integrates data from various sources, each subject to periodic updates, corrections, methodological improvements and changes in availability. For example, to measure industrial pollution of air, water and soil, we use data collected by the U.S. Environmental Protection Agency (EPA) on hundreds of toxic chemicals released from facilities across the U.S. This data is updated retroactively, filling in gaps or correcting past values. The model used to estimate area-level exposure to chemicals is also improved over successive releases. Each COI release uses the most recent EPA data release to provide the most accurate estimates of current and historical exposure to industrial pollutants. 

Second, our approach to standardizing component indicators and creating opportunity metrics considers all available data years and benchmarks to the latest year. To ensure efficient and robust parameters, component indicators are standardized using means and standard deviations computed over all available years of data. Opportunity metrics (Child Opportunity Levels and Scores) are benchmarked to the most recent year to be most instructive for contemporary analyses. 

Third, we continually review current and novel data sources, follow the rapidly evolving scientific research base and explore new technologies. We will update COI 3.0 through annual releases, and we also aim to develop and launch a new version of the COI once we reach a critical mass of substantial innovations that justifies a new version. Until then, we will filter small improvements in measurement quality into new releases. For example, with COI 3.0-2023, we modified our protocol for measuring healthy food access, taking advantage of new AI tools which incrementally improved predictive validity for measuring healthy food access.

Fourth, outside planned updates, new releases are necessary to correct errors in published data. For example, we identified and corrected an error in the Child Opportunity Index 2.0 soon after its launch and republished a corrected version of the index.

COI 3.0-2023 release notes

July 2025

For the most part, the COI 3.0-2023 follows the same methodologies and protocols as COI 3.0-2021 but also introduces the following small methodological changes. 

• Child Opportunity Levels (from “very low” to “very high”) and Child Opportunity Scores (1-100) benchmarks: Levels and Scores have been benchmarked to the 2023 neighborhood opportunity and population distributions. For example, for COI 3.0-2023, Child Opportunity Levels are constructed to contain 20% of the 2023 child population at each of the five Levels. COI 3.0-2021 metrics were benchmarked to the 2021 opportunity and population distribution and constructed to contain 20% of the 2021 child population at each of the five Levels.
• New releases for modeled school and pollution data: COI 3.0-2023 utilizes the Stanford Education Data Archive (SEDA) version 5.0, which extends school-level standardized test score estimates through the 2018/19 school year. COI 3.0-2021 used SEDA version 4.1, with data through 2017/18. COI 3.0-2023 incorporates the latest 2022 Risk-Screening Environmental Indicators (RSEI) estimates release, while COI 3.0-2021 used the 2020 RSEI release for the measurement of industrial pollutants in air, water and soil. For both SEDA and RSEI, the most recent releases include the most recent data, but they are not identical to previous releases. Historical data on pollution emitters or schools may have changed, and/or the model parameters used to obtain the modeled estimates have been updated between releases.
• COI 3.0-2023 uses a harmonized version of IRS Business Master Files for data on non-profits (501c3s), created by the National Center for Charitable Statistics (NCCS) at the Urban Institute. This single file replaces the multiple yearly files previously published by NCCS, which are no longer updated.
• Updated geographic definitions: COI 3.0-2023 uses county-equivalent planning regions for Connecticut, following Census Bureau conventions (see Appendix 6) and employs 2023 Office of Management and Budget (OMB) metropolitan area definitions. The transition to 2023 metro area definitions and changes in metro area population sizes have slightly altered which metro areas rank among the 100 largest.
• Updated standardization period: COI component indicators are standardized over all available years. Thus, COI 3.0-2023 component indicators are standardized using data over the  period from 2012 to 2023, while COI 3.0-2021 indicators were standardized over 2012 to 2021.
• Improved methodology for measuring healthy food retailer density: The approach to classifying healthy food retailers has been revised. COI 3.0-2021 combined text embeddings generated with OpenAI models with a traditional machine learning classification model. COI 3.0-2023 uses prompt-based few-shot classification of healthy food retailers.
• Income and earnings: Income and earnings indicators are deflated to 2023 U.S. dollars, whereas the COI 3.0-2021 data was deflated to 2021 U.S. dollars.
• Technical Document: The technical documentation has been fully updated to reflect the most recent data included in the current release. Older versions of the technical documentation continue to be available.

COI 3.0-2021 release notes

March 2024

COI 3.0-2021 was the first release of the Child Opportunity Index 3.0, replacing the 2.0 version. A summary of innovations is available in our blog What’s new in the Child Opportunity Index 3.0. COI 3.0-2021 continues to be available through our website, but it is no longer updated, and it does not contain the most recent data. Individuals and projects using the COI for the first time should use the COI 3.0-2023 release.

COI 2.0 Release Notes

June 2020

Due to an internal data processing error, two indicator z-scores were not reversed, i.e., multiplied by (-1) as described on page 15 of the Technical Document. The indicators affected are “hazardous waste dump sites” and “industrial pollutants in air, water or soil.” Correcting this error had downstream effects that marginally altered indicator weights in the health and environment domain, the health and environment domain overall average z-scores, domain average z-score weights and the overall average z-score. The impact on overall Child Opportunity metrics is marginal, however, because the indicators affected count relatively little within the affected domain. Initially published and revised overall z-scores are very highly correlated. Initially published and revised overall Child Opportunity Scores are also very highly correlated (Pearson’s rho = 0.99, national, state, and metro normed scores). After the data were recalculated, a small percentage of children were assigned to a different overall Child Opportunity Level: 4.9% (nationally normed), 5.0% (state normed) and 2.7% (metro normed). These reassignments occurred across adjacent Levels, e.g., from low to moderate, or very high to high, reflecting small changes in z-scores which shifted the opportunity Levels of only those neighborhoods with scores on the cusp of a Level boundary. 

We published revised versions of the COI 2.0 raw indicator file. In the original file, all variables were top- and bottom-coded, in accordance with our data processing protocols. However, for greater transparency, the revised data file includes variables without top and bottom codes. Moreover, the original file only included census tracts for which COI 2.0 metrics were calculated (see page 14 of the Technical Document). The revised data file contains all 2010 census tracts, including those with missing values.

We published revised versions of the child population data files. The previously published version only included census tracts for which COI 2.0 metrics were calculated (see page 14 of the Technical Document for details). The revised data file contains all 2010 census tracts. 

We published revised versions of the COI 2.0 component indicators z-scores. The revised data file contains additional variables, namely the indicator z-scores used to construct the economic resource index that is part of the social and economic domain. 

We published replication materials on our Github page. Using Stata 14 or higher, users can now replicate calculation of z-scores, domain and overall average z-scores, as well as Child Opportunity Scores and Child Opportunity Levels.

January 2020

COI 2.0 was released in January 2020. It was the first version of the Child Opportunity Index containing data for every neighborhood (census tract) in the U.S. and the first version to contain nationally-, state- and metro-normed data. It contains data from 2012 to 2017. COI 2.0 is available on our website but no longer updated. A summary of innovations and findings from COI 2.0 can be found in our press release, and additional information is available in the COI 2.0 Technical Document.